Bio-medical Ontologies Maintenance and Change Management

322 M.T. Hoque, M. Chetty, and A. Sattar
operation per seconds). This is still however, many orders of magnitude lower
than the requirement for a realistic solution.
With the objective of successfully building an effective computational strategy
to unravel the complexities of the sequence-to-folding relationship, even using the
well-established HP model, an efficient and robust solution has still to be developed.
In highlighting the various computational intelligence approaches for ab initio
PSP, the next section focuses mainly upon the low resolution HP model.
The HP Model
The HP model introduced by Dill [32, 33] is based on the fact that the hydrophobic
interactions dominate protein folding. The Hs form the protein core freeing up
energy, while the Ps, have an affinity with the solvent and so tend to remain in the
outer surface. For PSP, protein conformations of the sequence are placed as a selfavoiding
walk (SAW) on a 2D or 3D lattice. The energy of a given conformation
is defined as a number of topological neighbouring (TN) contacts between those
Hs, which are not sequential with respect to the sequence.
PSP is formally defined as: for an amino-acid sequence n s s s s s , , , , = 1 2 3 L , a
*
conformation c needs to be formed whereby c ∈ C(
s)
, energy
*
{ E(
c)
c C}
E = E(
C)
= min | ∈ [42], where n is the total amino acids in the sequence
and C (s)
is the set of all valid (i.e., SAW) conformations of s. If the number
of TNs in a conformation c is q then the value of E (c)
is defined as E( c)
= −q
and the fitness function is F = − q . The optimum conformation will have maximum
possible value of |F|. In a 2D HP square lattice model (Fig. 3. (a)), a nonterminal
and a terminal residue, both having 4 neighbours can have a maximum of
2 TNs and 3 TNs respectively. In a 2D FCC HP model (Fig. 3. (b)), a non-terminal
and a terminal residue both having 6 neighbours can have a maximum of 4 TNs
and 5 TNs respectively.
Many of the successful PSP software such as ROSETTA [4, 43], PROTINFO
[44, 45], TASSER [46] use various resolution of models embedded into the
(a) (b)
Fig. 3. Conformations in the 2D HP model shown by a solid line. (a) 2D square lattice having
fitness = - (TN Count) = -9. (b) 2D FCC lattice having fitness = -15. ‘ ’ indicates a
hydrophobic and ‘ ’ a hydrophilic residue. The dotted line indicates a TN. Starting residue
is indicated by ‘1’.